Add window-bounded String and char overloads to SubSequence

internal-api/src/main/java/datadog/trace/util/SubSequence.java

@@ -1,5 +1,7 @@
 package datadog.trace.util;
 
+import edu.umd.cs.findbugs.annotations.SuppressFBWarnings;
+
 /**
  * A <code>CharSequence</code> that is a view into a sub-sequence of a <code>String</code>. Unlike
  * <code>String.subSequence</code>, this class doesn't allocate an additional <code>String</code>,
@@ -66,8 +68,10 @@ public int length() {
 
   @Override
   public SubSequence subSequence(int start, int end) {
+    // start/end are offsets in THIS view's coordinates (CharSequence contract), so the absolute
+    // end is beginIndex + end -- NOT beginIndex + start + end (which overshoots by `start`).
     int newBeginIndex = this.beginIndex + start;
-    int newEndIndex = this.beginIndex + start + end;
+    int newEndIndex = this.beginIndex + end;
 
     return new SubSequence(this.str, newBeginIndex, newEndIndex);
   }
@@ -81,71 +85,169 @@ public void appendTo(StringBuilder builder) {
     if (beginIndex != endIndex) builder.append(this.str, beginIndex, endIndex);
   }
 
-  /** Returns the hash code as <code>backingStr.substr(beginIndex, endIndex).hashCode()</code> */
+  /**
+   * The same value as {@code toString().hashCode()} -- the {@link String} hash polynomial over this
+   * window -- but computed directly over the backing characters so hashing a view does not
+   * materialize a substring. Stays consistent with {@link #equals}: a view, its content-equal
+   * {@code String}, and an equal-content view all share this hash.
+   */
   @Override
   public int hashCode() {
-    return this.toString().hashCode();
+    int h = 0;
+    for (int i = this.beginIndex; i < this.endIndex; ++i) {
+      h = 31 * h + this.str.charAt(i);
+    }
+    return h;
   }
 
   /**
-   * Also handles String comparisons this.equals(backingStr.substr(beginIndex, endIndex)) is true
+   * Dispatches on the argument's runtime type: a {@code String} takes the {@link #equals(String)}
+   * region-compare fast path; any other {@code CharSequence} is compared via {@link
+   * #contentEquals(CharSequence)}. So {@code this.equals(backingStr.substring(beginIndex,
+   * endIndex))} is true, and two views with equal content are equal.
    */
+  // Intentionally asymmetric: a SubSequence equals any String/CharSequence with the same content,
+  // so subSeq.equals(str) can be true while str.equals(subSeq) is false. This view-equality
+  // convenience is by design (see class javadoc); accepted despite the equals() symmetry contract.
   @Override
+  @SuppressFBWarnings("EQ_CHECK_FOR_OPERAND_NOT_COMPATIBLE_WITH_THIS")
   public boolean equals(Object obj) {
-    if (!(obj instanceof CharSequence)) return false;
+    if (obj instanceof String) return this.equals((String) obj);
+    if (obj instanceof CharSequence) return this.contentEquals((CharSequence) obj);
 
-    return this.equals((CharSequence) obj);
+    return false;
   }
 
-  public final boolean equals(CharSequence that) {
-    int thisLen = this.length();
-    int thatLen = that.length();
+  /**
+   * Equivalent to {@code toString().equals(other)}. A thin alias for {@link
+   * #contentEquals(String)}, kept for callers working in terms of {@code equals}; prefer {@link
+   * #contentEquals(String)} directly.
+   */
+  public final boolean equals(String other) {
+    return this.contentEquals(other);
+  }
 
-    if (thisLen != thatLen) return false;
+  /**
+   * Equivalent to {@code toString().contentEquals(that)}: true when {@code that} has the same
+   * length and characters as this window. The general char-by-char comparison for any {@code
+   * CharSequence}; prefer {@link #contentEquals(String)} when the argument is known to be a {@code
+   * String}.
+   */
+  public final boolean contentEquals(CharSequence that) {
+    if (that == null) return false;
+
+    int len = this.length();
+    if (len != that.length()) return false;
 
-    for (int i = 0; i < Math.min(this.length(), that.length()); ++i) {
+    for (int i = 0; i < len; ++i) {
       if (this.charAt(i) != that.charAt(i)) return false;
     }
     return true;
   }
 
   /**
-   * True if this sub-sequence contains {@code needle} -- the zero-copy equivalent of {@code
-   * toString().contains(needle)}, with no substring materialized.
+   * String-specialized {@link #contentEquals(CharSequence)}: true when {@code other} has the same
+   * length and characters as this window, via {@link String#regionMatches} (backing-array compare,
+   * no per-char loop). Prefer this to {@link #equals(Object)} -- it states content-comparison
+   * intent and is free of the {@code equals} contract's cross-type asymmetry.
    */
-  public final boolean contains(String needle) {
-    return Strings.regionContains(this.str, this.beginIndex, this.endIndex, needle);
+  public final boolean contentEquals(String other) {
+    return other != null
+        && other.length() == this.length()
+        && this.str.regionMatches(this.beginIndex, other, 0, other.length());
   }
 
-  /** Case-insensitive content comparison; mirrors {@link String#equalsIgnoreCase(String)}. */
-  public final boolean equalsIgnoreCase(CharSequence that) {
-    int len = this.length();
-    if (that == null || len != that.length()) return false;
+  /**
+   * Case-insensitive counterpart of {@link #equals(String)}. Like {@link
+   * String#equalsIgnoreCase(String)}, a {@code null} argument is {@code false} rather than an
+   * error.
+   */
+  public final boolean equalsIgnoreCase(String other) {
+    return other != null
+        && other.length() == this.length()
+        && this.str.regionMatches(true, this.beginIndex, other, 0, other.length());
+  }
 
-    for (int i = 0; i < len; ++i) {
-      char a = this.charAt(i);
-      char b = that.charAt(i);
-      if (a != b) {
-        // Same two-way fold String.regionMatches(ignoreCase) uses (handles locale edge cases).
-        char au = Character.toUpperCase(a);
-        char bu = Character.toUpperCase(b);
-        if (au != bu && Character.toLowerCase(au) != Character.toLowerCase(bu)) {
-          return false;
-        }
-      }
-    }
-    return true;
+  /**
+   * Equivalent to {@code toString().startsWith(prefix)}. The window guard ({@code prefix.length()
+   * <= length()}) keeps the delegated read inside {@code [beginIndex, endIndex)}.
+   */
+  public final boolean startsWith(String prefix) {
+    return prefix.length() <= this.length() && this.str.startsWith(prefix, this.beginIndex);
   }
 
-  /** True if this sub-sequence begins with {@code prefix} (content comparison, no allocation). */
-  public final boolean startsWith(CharSequence prefix) {
-    int prefixLen = prefix.length();
-    if (prefixLen > this.length()) return false;
+  /**
+   * Equivalent to {@code length() > 0 && charAt(0) == c}, the single-character {@link
+   * #startsWith(String)}.
+   */
+  public final boolean startsWith(char c) {
+    return this.beginIndex < this.endIndex && this.str.charAt(this.beginIndex) == c;
+  }
 
-    for (int i = 0; i < prefixLen; ++i) {
-      if (this.charAt(i) != prefix.charAt(i)) return false;
-    }
-    return true;
+  /**
+   * Equivalent to {@code toString().endsWith(suffix)}. Implemented as a prefix match anchored at
+   * {@code endIndex - suffix.length()} so the read stays inside this window.
+   */
+  public final boolean endsWith(String suffix) {
+    int suffixLen = suffix.length();
+    return suffixLen <= this.length() && this.str.startsWith(suffix, this.endIndex - suffixLen);
+  }
+
+  /**
+   * Equivalent to {@code length() > 0 && charAt(length() - 1) == c}, the single-character {@link
+   * #endsWith(String)}.
+   */
+  public final boolean endsWith(char c) {
+    return this.beginIndex < this.endIndex && this.str.charAt(this.endIndex - 1) == c;
+  }
+
+  /**
+   * Equivalent to {@code toString().indexOf(needle)}: the offset of the first full occurrence of
+   * {@code needle} within this window relative to the window start, or {@code -1} if it does not
+   * occur fully in range. {@link String#indexOf(String, int)} returns the earliest occurrence at or
+   * after {@code beginIndex}, so a single bound check against {@code endIndex} is exact.
+   */
+  public final int indexOf(String needle) {
+    int idx = this.str.indexOf(needle, this.beginIndex);
+    return (idx >= 0 && idx + needle.length() <= this.endIndex) ? idx - this.beginIndex : -1;
+  }
+
+  /**
+   * Equivalent to {@code toString().indexOf(c)}: the offset of the first {@code c} within this
+   * window relative to the window start, or {@code -1} if it does not occur in range.
+   */
+  public final int indexOf(char c) {
+    int idx = this.str.indexOf(c, this.beginIndex);
+    return (idx >= 0 && idx < this.endIndex) ? idx - this.beginIndex : -1;
+  }
+
+  /**
+   * Equivalent to {@code toString().lastIndexOf(needle)}: the offset of the last full occurrence of
+   * {@code needle} within this window relative to the window start, or {@code -1} if it does not
+   * occur fully in range. Searches back from {@code endIndex - needle.length()} -- the latest start
+   * whose end still fits the window -- so the lower bound is a single check against {@code
+   * beginIndex}.
+   */
+  public final int lastIndexOf(String needle) {
+    int idx = this.str.lastIndexOf(needle, this.endIndex - needle.length());
+    return (idx >= this.beginIndex) ? idx - this.beginIndex : -1;
+  }
+
+  /**
+   * Equivalent to {@code toString().lastIndexOf(c)}: the offset of the last {@code c} within this
+   * window relative to the window start, or {@code -1} if it does not occur in range.
+   */
+  public final int lastIndexOf(char c) {
+    int idx = this.str.lastIndexOf(c, this.endIndex - 1);
+    return (idx >= this.beginIndex) ? idx - this.beginIndex : -1;
+  }
+
+  /**
+   * True if this sub-sequence contains {@code needle} -- the zero-copy equivalent of {@code
+   * toString().contains(needle)}, with no substring materialized.
+   */
+  public final boolean contains(String needle) {
+    return Strings.regionContains(this.str, this.beginIndex, this.endIndex, needle);
   }
 
   @Override

internal-api/src/test/java/datadog/trace/util/SubSequenceTest.java

@@ -125,27 +125,136 @@ public void contains() {
   }
 
   @Test
-  public void equalsIgnoreCase() {
-    SubSequence call = SubSequence.of("xx CALL yy", 3, 7); // "CALL"
+  public void subSequenceOfView() {
+    // Instance subSequence(start, end): start/end are in THIS view's coordinates (CharSequence
+    // contract), regardless of where the view sits in the backing string.
+    SubSequence view = SubSequence.of("abcdefghij", 2, 8); // "cdefgh"
+    SubSequence mid = view.subSequence(1, 4); // chars [1, 4) of "cdefgh" -> "def"
+    assertEquals("def", mid.toString());
+    assertEquals(3, mid.beginIndex()); // absolute begin = 2 + 1
+    assertEquals(6, mid.endIndex()); // absolute end = 2 + 4 (NOT 2 + 1 + 4)
+
+    // full window and empty are exact
+    assertEquals("cdefgh", view.subSequence(0, view.length()).toString());
+    assertEquals("", view.subSequence(2, 2).toString());
+
+    // nested: subSequence of a non-zero-start view stays correct (the case the old bug broke worst)
+    assertEquals("ef", mid.subSequence(1, 3).toString()); // chars [1, 3) of "def" -> "ef"
+  }
+
+  @Test
+  public void equalsString() {
+    // "call" sits at [6, 10) inside the backing string, flanked by other text.
+    SubSequence call = SubSequence.of("xxxxx call yyyyy", 6, 10);
+    assertTrue(call.equals("call"));
+    assertFalse(call.equals("CALL")); // case-sensitive
+    assertFalse(call.equals("cal")); // shorter
+    assertFalse(call.equals("calls")); // longer (would overshoot endIndex)
+    assertFalse(call.equals((Object) null));
+
+    // equals(Object) routes a String through the region-compare fast path...
+    assertTrue(call.equals((Object) "call"));
+    // ...and any other CharSequence (incl. another SubSequence) through contentEquals.
+    assertTrue(call.equals((Object) new StringBuilder("call")));
+    assertTrue(call.equals((Object) SubSequence.of("xxxxx call yyyyy", 6, 10)));
+    assertFalse(call.equals((Object) Integer.valueOf(4)));
+  }
+
+  @Test
+  public void contentEqualsCharSequence() {
+    SubSequence call = SubSequence.of("xxxxx call yyyyy", 6, 10); // "call"
+    assertTrue(call.contentEquals("call")); // String is a CharSequence
+    assertTrue(call.contentEquals(new StringBuilder("call")));
+    assertTrue(call.contentEquals(SubSequence.of("a call b", 2, 6)));
+    assertFalse(call.contentEquals("CALL")); // case-sensitive
+    assertFalse(call.contentEquals("cal")); // length mismatch
+    assertFalse(call.contentEquals(null));
+  }
+
+  @Test
+  public void equalsIgnoreCaseString() {
+    SubSequence call = SubSequence.of("xxxxx CaLl yyyyy", 6, 10);
     assertTrue(call.equalsIgnoreCase("call"));
     assertTrue(call.equalsIgnoreCase("CALL"));
-    assertTrue(call.equalsIgnoreCase("CaLl"));
-    assertFalse(call.equalsIgnoreCase("calls")); // length differs
-    assertFalse(call.equalsIgnoreCase("cant")); // same length, content differs
+    assertFalse(call.equalsIgnoreCase("cal"));
+    assertFalse(call.equalsIgnoreCase("calls"));
+    assertFalse(call.equalsIgnoreCase(null)); // matches String.equalsIgnoreCase(null)
+  }
+
+  @Test
+  public void startsWithString() {
+    SubSequence view = SubSequence.of("xx{call}xx", 2, 8); // "{call}"
+    assertTrue(view.startsWith("{"));
+    assertTrue(view.startsWith("{call"));
+    assertTrue(view.startsWith("{call}"));
+    assertFalse(view.startsWith("call"));
+    assertFalse(view.startsWith("{call}x")); // overshoots endIndex even though backing has 'x'
+    assertTrue(view.startsWith("")); // empty prefix
+  }
 
-    // case-sensitive equals stays case-sensitive
-    assertFalse(call.equals("call"));
-    assertTrue(call.equals("CALL"));
+  @Test
+  public void endsWithString() {
+    SubSequence view = SubSequence.of("xx{call}xx", 2, 8); // "{call}"
+    assertTrue(view.endsWith("}"));
+    assertTrue(view.endsWith("call}"));
+    assertTrue(view.endsWith("{call}"));
+    assertFalse(view.endsWith("call"));
+    assertFalse(view.endsWith("x{call}")); // undershoots beginIndex even though backing has 'x'
+    assertTrue(view.endsWith("")); // empty suffix
+  }
+
+  @Test
+  public void indexOfString() {
+    SubSequence view = SubSequence.of("aa-bc-bc-aa", 3, 8); // "bc-bc"
+    assertEquals(0, view.indexOf("bc")); // window-relative offset of the first occurrence
+    assertEquals(2, view.indexOf("-bc")); // non-zero relative offset
+    assertEquals(2, view.indexOf("-"));
+    assertEquals(-1, view.indexOf("aa")); // present in backing string but outside the window
+    assertEquals(-1, view.indexOf("bc-bc-")); // overshoots endIndex
+  }
+
+  @Test
+  public void lastIndexOfString() {
+    SubSequence view = SubSequence.of("aa-bc-bc-aa", 3, 8); // "bc-bc"
+    assertEquals(3, view.lastIndexOf("bc")); // last "bc" -> relative 3
+    assertEquals(2, view.lastIndexOf("-"));
+    assertEquals(-1, view.lastIndexOf("aa")); // outside the window on both ends
+    assertEquals(-1, view.lastIndexOf("bc-bc-")); // overshoots endIndex
+  }
+
+  @Test
+  public void startsWithChar() {
+    SubSequence view = SubSequence.of("xx{call}xx", 2, 8); // "{call}"
+    assertTrue(view.startsWith('{'));
+    assertFalse(view.startsWith('c')); // 'c' is at offset 1, not the start
+    assertFalse(view.startsWith('x')); // backing char before beginIndex, outside the window
+    assertFalse(SubSequence.EMPTY.startsWith('x')); // empty window
+  }
+
+  @Test
+  public void endsWithChar() {
+    SubSequence view = SubSequence.of("xx{call}xx", 2, 8); // "{call}"
+    assertTrue(view.endsWith('}'));
+    assertFalse(view.endsWith('l')); // 'l' is one before the end
+    assertFalse(view.endsWith('x')); // backing char at endIndex, outside the window
+    assertFalse(SubSequence.EMPTY.endsWith('x')); // empty window
+  }
+
+  @Test
+  public void indexOfChar() {
+    SubSequence view = SubSequence.of("aa-bc-bc-aa", 3, 8); // "bc-bc"
+    assertEquals(0, view.indexOf('b')); // window-relative offset of the first occurrence
+    assertEquals(1, view.indexOf('c'));
+    assertEquals(2, view.indexOf('-'));
+    assertEquals(-1, view.indexOf('a')); // present in backing string but outside the window
   }
 
   @Test
-  public void startsWith() {
-    SubSequence braceCall = SubSequence.of("xx{call} yy", 2, 7); // "{call"
-    assertTrue(braceCall.startsWith(""));
-    assertTrue(braceCall.startsWith("{"));
-    assertTrue(braceCall.startsWith("{ca"));
-    assertTrue(braceCall.startsWith("{call"));
-    assertFalse(braceCall.startsWith("call")); // not the prefix
-    assertFalse(braceCall.startsWith("{calls and more")); // prefix longer than sequence
+  public void lastIndexOfChar() {
+    SubSequence view = SubSequence.of("aa-bc-bc-aa", 3, 8); // "bc-bc"
+    assertEquals(3, view.lastIndexOf('b')); // last 'b' -> relative 3
+    assertEquals(4, view.lastIndexOf('c'));
+    assertEquals(2, view.lastIndexOf('-'));
+    assertEquals(-1, view.lastIndexOf('a')); // outside the window on both ends
   }
 }